How to serve a WSGI application via CGI
CGI has a bad reputation. Its slow performance disqualify it for most demanding tasks; nobody would think of serving thousands of pages per seconds via CGI. But when performance doesn’t matter, it can still be useful.
CGI is easy to deploy, most HTTP servers fully support it, and writing a minimal script only takes a few lines:
print 'Content-type: text/plain'
print
print 'Hello World!'
But writing CGI scripts in Python can be difficult. Unlike PHP where most “web” features are embedded in the language. Python CGI scripts don’t handle query strings, path info, cookies, sessions, etc. out of the box. They usually rely on the cgi module and some custom code to handle this.
WSGI also can provide those missing features. There are lots of existing middlewares and utilities for WSGI providing all kinds of services. Also the code developed for your CGI scripts can be reused in other applications, even those using SCGI or FastCGI. wsgiref is shipped with CPython since version 2.5, and a CGI handler is included; no need to install extra packages to get it running.
To run your WSGI application pass it to the run method of the CGIHandler:
#!/usr/bin/env python
import wsgiref.handlers
def application(environ, start_response):
start_response('200 OK', [('Content-Type', 'text/plain')])
return ['Hello World!\n']
if __name__ == '__main__':
wsgiref.handlers.CGIHandler().run(application)
You might have to configure your HTTP server to serve CGI scripts. Setting it up is usually a lot simpler than FastCGI, SCGI, mod_python, and mod_wsgi:
To deploy your scripts: copy them to the appropriate directory, that’s all, no need to restart a server. CGI is appealing for small scripts because of its easy deployment. Remember it next time you have to do “this simple script that shouldn’t take long to develop”, often deployment time accounts for more than development time.